DM-16467 added some Python and bash scripts that add brighter-fatter kernel and defect datasets to a Gen3 repo.  These use direct SQL calls instead of the Registry API, and have been broken by changes on DM-16227.  This ticket will move that work into the daf.butler.Instrument class and its obs_* package specializations, allowing it to be done automatically by the gen2convert tool (in addition to fixing the breakage, of course).

Coming up with a new, consistent-across-obs-packages way to manage human-curated source for these datasets is definitely out of scope right now, but this might provide an example of how the Butler side of that might look to those who do design that.